Running Back Workload

Conventional wisdom suggests that, all else being equal, running backs with "low mileage" are more likely to age gracefully than running backs who have accumulated a significant number of carries.

This, unfortunately, is a very complicated issue to test. For example, new Giants running back Rashad Jennings is 29 years old, but he has just 387 career carries.  This makes Jennings a "young" 29, but is that better than being an "old" 28? The best way to test this question is to analyze running backs of similar quality as Jennings -- but who had a lot of carries by the time they were 28 years old -- and see how the rest of their careers unfolded.  The problem is that the list of running backs with a lot of carries through their age 28 season bear no resemblance to Jennings. The players with the most carries through age 28 are Emmitt Smith, Edgerrin James, Jerome Bettis, Barry Sanders, LaDainian Tomlinson, Curtis Martin, and Walter Payton, which basically serves as a who's who of running backs who are not comparable to Rashad Jennings.

Generally speaking, the best running backs get the most carries: did you know that Jim Brown is the only player to lead the NFL in carries more than 4 times? He did it six times in his nine-year career. Along the same line of thinking, the running backs with the most carries are generally among the best running backs.  Running backs who haven't had a lot of carries through age 28 generally either aren't very good or have suffered multiple injuries, which makes it tough to find players who feel like true comparables to a player like Jennings.

One could argue that running back workload and running back quality are so inextricably tied that it's impossible to accurately measure whether age or workload is more important.  But today, I want to take a step back from examining the specifics of a player like Jennings and look at the big picture.  There are some examples that appear to support the "running back mileage" theory.  Shaun Alexander had a significant number of carries through age 28, and was excellent at age 28; the fact that he then declined so significantly, so quickly, could be a sign that workload really mattered. After all, few players suffer such sharp declines when turning 29. But that's just one data point.  What if we can bring in many more?

I looked at the 40 running backs with the most career VBD to enter the league since 1988 and that have since retired.  Then, for each player at each age of his career, I calculated his:

• Age;
• Carries at that age
• Career carries through that age
• VBD at that age
• VBD through that age
• Future VBD

In other words, here is how Shaun Alexander's line would read following his age 28 season:

At age 28, Alexander had a monster season, recording 370 carries (and bringing his number of career carries up to 1,717) and producing 211.8 points of VBD (and bringing his total number of VBD to 761). But Alexander, who won the Associated Press NFL MVP award that year, never produced another season with positive VBD. That's pretty crazy, and goes down as an example of running back overuse.

But, again, that's just one example.  I performed a regression analysis using those first five variables -- age, rush attempts at that age, career rushing attempts through that age, VBD at that age, and career VBD through that age -- to predict a sixth variable: future VBD.  Here was the best-fit formula using all age 27+ seasons for the 40 running backs determined above:

Future VBD = 492 -14.98 * Age - 0.021 * Rush_Att - 0.01 * Car_Rush_Att + 0.653 * VBD + 0.004 * Car_VBD

Now that formula might look like gibberish to some of you.  But the number we are concerned about is in green: -0.01. That's the amount of weight that should be placed on the "career rushing attempts through age X" variable. You don't need a calculator to know that -0.01 is very close to zero. Meanwhile, the "-14.98" coefficient blue tells us that for every year, we should project about 15 fewer points of future VBD for a player.  If you look at the coefficients on the age and career rush attempts variables, it tells us that being one year older is like having endured an additional 1500 carries: in other words, this suggests that age is much, much, much more important than workload. However, there are some reasons to be skeptical of this result, which we'll get to at the end.

In general, you might find that these results might seem a bit counterintuitive.  After all, what about Shaun Alexander? Let's use these numbers to "project" Alexander after his age 28 season.  We would begin with 492, subtract (14.98 * 28), subtract (0.021 * 370), subtract (0.01 * 1717), add (0.653 * 211.8), and add (0.004 * 761): that gives him a projection of 189 points of future VBD. That, obviously, was very wrong. 

As it turns out, there were 206 seasons worth of "projections" that this regression analyzed. And Alexander's age 29 season was the most extreme underachieving projection, so using Alexander as an example of how players age is not very wise. In fact, I'll post all 206 seasons for those who are curious:

Let's take a closer look at this analysis, and see if we can identify any problems.

• Tied with Alexander for the "biggest underachiever" was Larry Johnson.  But Johnson goes down in this analysis as a "low-mileage" player. Does that seem right to you?  He had under 900 career carries, so one certainly could have put forth the "he's younger than his age" argument. On the other hand, many feel that his 400+ carry season (followed by a lengthy holdout).  So is Johnson an example for or against the mileage argument?  This is the sort of ambiguity in the debate that often gets lost. Maybe the takeaway from this regression is that workload doesn't matter, or maybe it just means that we're measuring it improperly. Another possibility: the effects of workload vary drastically from player to player, and are simply too subtle to have been picked up in these studies.

• Priest Holmes is often cited as an example of why we should focus on mileage: He had just 459 carries through age 27, and then exceeded expectations for several years.  But age did catch up to him: entering his age 30 season, fantasy owners used the "low mileage" argument to say that Holmes was a young 30. Unfortunately, injuries began to ruin his career that season.  Is that an argument for age in favor of mileage, or was it just random variation? Who knows. But we know that being a "young 30" in the summer of 2004 didn't help the folks who drafted Holmes that year. Another explanation: after three straight years of 300+ carries, he wasn't so "young" anymore.

• Dorsey Levens could have been argued as a "young" 30 entering the 2000 season, but again, injuries quickly put an end to his days as a fantasy stud. Then again, Levens had a spotty injury history, which is one of the reasons why he didn't log so many carries early in his career. As you can see, figuring out who is "young" and who is "old" can be pretty complicated.

The biggest overachievers include some "low mileage" backs like Holmes and Tiki Barber... but it also includes Barry Sanders, Curtis Martin, and Emmitt Smith. Those players continued to rack up touches year after year, and they continued to defy any expectations of regression based on heavy workload.  It's easy to say "hey, those are four outstanding running backs, you can't compare regular backs to them"... but we only know that they were outstanding because they continued to produce year after year. 

Over the last three years, Marshawn Lynch leads the NFL with 1,002 carries (including the postseason).  If he is this decade's version of Alexander in Seattle, you can be sure that the narrative will be that a heavy workload prematurely ended his career.  But if Lynch puts together three more seasons of strong production, the narrative will be: sure, workload matters, unless you're a special back like Marshawn Lynch! Narratives are fun, but they're not helpul in predicting the future. If Shaun Alexander or Maurice Jones-Drew maintained their levels into their late 20s and early 30s, they would be Hall of Fame backs, too.  And if you're looking for a non-HOFer, Ricky Watters also managed to maintain his fantasy value despite accumulating significant touches.

So does workload not matter?  Certainly one interpretation of the analysis here is that workload is not nearly as important as people think.  But we need to remember the parameters of this study: we are looking only at the 40 top fantasy running backs over roughly a 20-year period. There is some survivorship bias in the data, as running backs who faded before they accumulated enough fantasy points to qualify would be ignored here (although I'm not sure if we would say they declined due to workload, but that's for another post).  In addition, because we have a closed sample of backs, the amount of "future VBD" for each back is necessarily going to be limited. In Part II, I'll attempt to address these issues.